Resource allocation method and device, electronic device and storage medium

Abstract

The embodiments of the present application provide a resource allocation method and apparatus, an electronic device, and a storage medium, belonging to the field of artificial intelligence technology. The method comprises: obtaining resource transaction configuration information of a target resource, obtaining historical transaction data of the target resource; obtaining a candidate reward model based on the resource transaction configuration information; performing a genetic search based on the candidate reward model and historical transaction data to obtain a target reward function; training a preset decision generation model based on the target reward function and historical transaction data to obtain a target decision generation model; obtaining current resource transaction data of the target resource, and making an allocation decision on the current resource transaction data based on the target decision generation model to obtain target resource allocation data. The embodiments of the present application can improve the efficiency and accuracy of resource allocation.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a resource allocation method and device, an electronic device, and a storage medium. Background Art

[0002] Resource allocation can involve allocating production materials, equipment, funds, computer software, and hardware resources. For example, it can allocate medical device inventory. Because the value of medical device inventory is subject to market fluctuations and the carrying costs associated with inventory lead to the risk of losses, timely allocation of medical device inventory is crucial. Current resource allocation methods primarily rely on manual inventory counting and allocation, resulting in low efficiency and prone to human error. Therefore, improving the efficiency and accuracy of resource allocation has become a pressing issue. Summary of the Invention

[0003] The main purpose of the embodiments of the present application is to propose a resource allocation method and device, an electronic device and a storage medium, aiming to improve the efficiency and accuracy of resource allocation.

[0004] To achieve the above objectives, a first aspect of an embodiment of the present application provides a resource allocation method, the method comprising:

[0005] Obtain resource transaction configuration information of the target resource and obtain historical transaction data of the target resource;

[0006] Acquire a candidate reward model based on the resource transaction configuration information;

[0007] Performing a genetic search based on the candidate reward model and the historical transaction data to obtain a target reward function;

[0008] Training a preset decision generation model based on the target reward function and the historical transaction data to obtain a target decision generation model;

[0009] Current resource transaction data of the target resource is acquired, and allocation decision is made on the current resource transaction data based on the target decision generation model to obtain target resource allocation data.

[0010] In some embodiments, performing a genetic search based on the candidate reward model and the historical transaction data to obtain a target reward function includes:

[0011] Perform chromosome encoding on the candidate reward model to obtain an initial function population;

[0012] Performing fitness screening on the initial function population according to the historical transaction data to obtain a first function population;

[0013] Performing crossover and mutation processing on the initial function population to obtain a second function population;

[0014] Performing a local search based on the second function population and the historical transaction data to obtain a third function population;

[0015] Performing population fitness replacement on the second function population according to the first function population to obtain a fourth function population;

[0016] Merging the third function population and the fourth function population to obtain a merged function population;

[0017] Fitness screening is performed on the merged function population according to the historical transaction data to obtain the target reward function.

[0018] In some embodiments, the step of performing fitness screening on the combined function population according to the historical transaction data to obtain the target reward function includes:

[0019] Calculating the reward value of the merged function population according to the historical transaction data to obtain target reward value data;

[0020] Allocate resources to the historical transaction data according to the reward value data to obtain resource income data and resource withdrawal value data;

[0021] Performing fitness calculation based on the resource income data and the resource withdrawal value data to obtain population fitness data;

[0022] The merged function population is screened according to the population fitness data to obtain the target reward function.

[0023] In some embodiments, the calculating of the reward value based on the combined function population and the historical transaction data to obtain target reward value data includes:

[0024] Calculating a resource retention reward value for the merged function population according to the historical transaction data to obtain resource retention reward value data;

[0025] Calculating the resource output reward value of the merged function population according to the historical transaction data to obtain resource output reward value data;

[0026] Calculating the resource input reward value for the merged function population according to the historical transaction data to obtain resource input reward value data;

[0027] The resource retention bonus value data, the resource output bonus value data, and the resource input bonus value data are used as the target bonus value data.

[0028] In some embodiments, the preset decision generation model includes experience samples, an initial decision layer, and an initial prediction layer; the experience samples include an executed action, a current state, and a next state; and training the preset decision generation model based on the target reward function and the historical transaction data to obtain a target decision generation model includes:

[0029] Performing a decision evaluation based on the execution action, the current state, and the initial decision layer to obtain initial decision evaluation data; wherein the initial decision evaluation data is used to indicate the rationality of the decision of the initial decision layer;

[0030] Performing a prediction evaluation based on the execution action, the next state, and the initial prediction layer to obtain initial prediction evaluation data; wherein the initial prediction evaluation data is used to indicate the rationality of the decision of the initial prediction layer;

[0031] Calculating a reward value based on the historical transaction data, the current state, the executed action, the next state, and the target reward function to obtain a training reward value;

[0032] Optimizing parameters of the initial decision layer according to the training reward value, the initial prediction evaluation data, and the initial decision evaluation data to obtain a target decision layer;

[0033] The target decision generation model is formed according to the experience samples, the target decision layer and the initial prediction layer.

[0034] In some embodiments, the step of optimizing parameters of the initial decision layer according to the training reward value, the initial prediction evaluation data, and the initial decision evaluation data to obtain a target decision layer includes:

[0035] Performing aggregate calculation based on the training reward value and the initial prediction evaluation data to obtain a training label value;

[0036] Calculate the mean square error based on the training label value and the initial decision evaluation data to obtain a training loss value;

[0037] Parameters of the initial decision layer are optimized according to the training loss value to obtain the target decision layer.

[0038] In some embodiments, obtaining a candidate reward model based on the resource transaction configuration information includes:

[0039] Perform semantic extraction on the resource transaction configuration information according to a preset semantic extraction model to obtain transaction loss data and transaction number limit data;

[0040] Constructing a loss model based on the transaction loss data to obtain a loss sub-model;

[0041] Constructing a restriction model based on the transaction number restriction data to obtain a restriction sub-model;

[0042] A model is constructed based on the loss sub-model, the restriction sub-model, the preset risk control sub-model and the preset benefit control sub-model to obtain the candidate reward model.

[0043] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a resource allocation device, the device comprising:

[0044] A data acquisition module is used to acquire resource transaction configuration information of a target resource and acquire historical transaction data of the target resource;

[0045] A model acquisition module, configured to acquire a candidate reward model based on the resource transaction configuration information;

[0046] A genetic search module, configured to perform a genetic search based on the candidate reward model and the historical transaction data to obtain a target reward function;

[0047] A model training module, configured to train a preset decision generation model based on the target reward function and the historical transaction data to obtain a target decision generation model;

[0048] The allocation decision module is used to obtain the current resource transaction data of the target resource and make an allocation decision on the current resource transaction data based on the target decision generation model to obtain the target resource allocation data.

[0049] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the method described in the first aspect when executing the computer program.

[0050] To achieve the above-mentioned purpose, the fourth aspect of the embodiments of the present application proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method described in the first aspect.

[0051] The present application proposes a resource allocation method and device, electronic device and storage medium, which obtain resource transaction configuration information and historical transaction data of the target resource, obtain a candidate reward model based on the resource transaction configuration information, and perform genetic search based on the reward function model and historical transaction data to obtain a target reward function. Then, the preset decision generation model is trained based on the target reward function and historical transaction data, so that the preset decision generation model can be trained into a target decision generation model that can make resource allocation decisions with high allocation efficiency under the reward incentive of the target reward function, thereby improving the efficiency and accuracy of resource allocation; further, the current resource transaction data of the target resource is obtained, and an allocation decision is made on the current resource transaction data based on the target decision generation model to obtain target resource allocation data, thereby achieving efficient and accurate resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a flow chart of the resource allocation method provided in an embodiment of the present application;

[0053] Figure 2 yes Figure 1 Flowchart of step S102 in FIG.

[0054] Figure 3 yes Figure 1 Flowchart of step S103 in FIG.

[0055] Figure 4 yes Figure 3 Flowchart of step S307 in FIG.

[0056] Figure 5 yes Figure 4 Flowchart of step S401 in FIG.

[0057] Figure 6 yes Figure 1 Flowchart of step S104 in FIG.

[0058] Figure 7 yes Figure 6 Flowchart of step S604 in FIG.

[0059] Figure 8 is a structural diagram of a resource allocation device provided in an embodiment of the present application;

[0060] Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0062] It should be noted that although the device schematics illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the device or the sequence in the flowcharts. The terms "first," "second," and so on, in the specification, claims, and drawings, are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

[0063] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0064] First, let’s analyze some of the terms used in this application:

[0065] Genetic search is a biomimetic optimization algorithm designed to simulate the process of biological evolution in nature to solve complex optimization problems. Based on the theory of genetic algorithms, genetic search uses operations such as selection, crossover, and mutation to iteratively find the optimal solution through population evolution. By selecting individuals with high fitness for reproduction, genetic search gradually improves the solutions in the population, ultimately converging on the optimal solution. Genetic search can handle a variety of optimization problems and is particularly well-suited for problems with high nonlinearity, complex constraints, or large search spaces. Key concepts involved in genetic search include fitness assessment, individual encoding, selection strategies, and crossover and mutation operations.

[0066] Deep Q-Network (DQN): An algorithm that combines deep learning and reinforcement learning to solve decision-making problems in complex environments. Based on the Q-learning algorithm, Deep Q-Network uses a deep neural network to approximate the Q-value function, evaluating the value of taking an action in a specific state. The goal of Deep Q-Network is to maximize cumulative rewards through training strategies. Deep Q-Networks have strong expressiveness in complex, high-dimensional state spaces. Deep Q-Networks utilize an experience replay mechanism to store and randomly sample the agent's interaction data, preventing data correlation from affecting training stability. Deep Q-Networks also introduce a target network to mitigate training instability.

[0067] The reward function is a core concept in reinforcement learning, used to measure the feedback or reward an agent receives after taking an action in the environment. The reward function provides the agent with immediate feedback on the quality of its actions, enabling it to adjust its strategy to maximize its cumulative reward. The reward function defines the goal within the environment and guides the agent towards the optimal solution. In reinforcement learning, the agent continuously learns through interaction with the environment. An action is considered optimal when it maximizes the reward function under a specific state.

[0068] The fitness function is a core concept in genetic algorithms and evolutionary computation, used to evaluate the performance of candidate solutions for a given problem. By quantifying the performance of each individual, the fitness function guides the algorithm to gradually optimize the solution set, mimicking the "survival of the fittest" process in nature. The fitness function's evaluation of an individual is typically given as a numerical value, representing the individual's fitness level. Individuals with high fitness are more likely to be selected for crossover and mutation, generating the next generation of solutions.

[0069] Resource allocation can involve allocating production materials, equipment, funds, computer software, and hardware resources. For example, it can allocate medical device inventory. Because the value of medical device inventory is subject to market fluctuations and the carrying costs associated with inventory lead to the risk of losses, timely allocation of medical device inventory is crucial. Current resource allocation methods primarily rely on manual inventory counting and allocation, resulting in low efficiency and prone to human error. Therefore, improving the efficiency and accuracy of resource allocation has become a pressing issue.

[0070] Based on this, embodiments of the present application provide a resource allocation method and apparatus, an electronic device, and a storage medium, aiming to improve the efficiency and accuracy of resource allocation.

[0071] The embodiments of the present application provide a resource allocation method and device, an electronic device, and a storage medium, which are specifically illustrated through the following embodiments. First, the resource allocation method in the embodiments of the present application is described.

[0072] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results.

[0073] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0074] The resource allocation method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The resource allocation method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or can be configured as a server cluster or a distributed system composed of multiple physical servers, or can be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the resource allocation method, etc., but is not limited to the above forms.

[0075] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0076] It should be noted that in each specific embodiment of the present application, when it comes to the need to perform relevant processing based on data related to the user's identity or characteristics, such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of such data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0077] Figure 1 This is an optional flowchart of the resource allocation method provided in the embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S105.

[0078] Step S101: obtaining resource transaction configuration information of a target resource and obtaining historical transaction data of the target resource;

[0079] Step S102, obtaining a candidate reward model based on resource transaction configuration information;

[0080] Step S103: performing a genetic search based on the candidate reward model and historical transaction data to obtain a target reward function;

[0081] Step S104: training a preset decision generation model based on the target reward function and historical transaction data to obtain a target decision generation model;

[0082] Step S105 , obtaining current resource transaction data of the target resource, and making an allocation decision on the current resource transaction data based on the target decision generation model to obtain target resource allocation data.

[0083] In steps S101 to S105 shown in the embodiment of the present application, resource transaction configuration information and historical transaction data of the target resource are obtained, a candidate reward model is obtained based on the resource transaction configuration information, and a genetic search is performed based on the reward function model and the historical transaction data to obtain a target reward function. Then, the preset decision generation model is trained based on the target reward function and the historical transaction data, so that the preset decision generation model can be trained into a target decision generation model that can make resource allocation decisions with high allocation efficiency under the reward incentive of the target reward function, thereby improving the efficiency and accuracy of resource allocation; further, the current resource transaction data of the target resource is obtained, and an allocation decision is made on the current resource transaction data based on the target decision generation model to obtain target resource allocation data, thereby achieving efficient and accurate resource allocation.

[0084] In step S101 of some embodiments, the target resource is an item, which can be an electronic item or an actual physical item, such as a medical device product in inventory. Since the product value of medical device products is affected by market fluctuations, and since there are holding costs for medical products in inventory, such as natural damage during storage or inventory warehousing fees, there is a risk of loss of medical device products in inventory, so it is necessary to allocate the medical device products in inventory in a timely manner. When allocating resources, there is resource transaction configuration information. Resource transaction configuration information refers to the rules for resource allocation of target resources, such as inventory shipment quantity restrictions, such as natural loss during transportation, etc. Historical transaction data refers to historical transaction value data of the target resource.

[0085] See also Figure 2 In some embodiments, step S102 may include but is not limited to steps S201 to S204:

[0086] Step S201: performing semantic extraction on resource transaction configuration information according to a preset semantic extraction model to obtain transaction loss data and transaction number limit data;

[0087] Step S202: constructing a loss model based on transaction loss data to obtain a loss sub-model;

[0088] Step S203, constructing a restriction model based on the transaction number restriction data to obtain a restriction sub-model;

[0089] Step S204 , constructing a model based on the loss sub-model, the restriction sub-model, the preset risk control sub-model, and the preset benefit control sub-model to obtain a candidate reward model.

[0090] In steps S201 to S204 shown in the embodiment of the present application, semantic extraction is performed on the resource transaction configuration information according to a preset semantic extraction model to obtain transaction loss data and transaction number limit data, thereby significantly reducing the time and effort required for manual operations, improving allocation efficiency and avoiding human errors. A loss model is then constructed based on the transaction loss data to obtain a loss sub-model. A restriction model is then constructed based on the transaction number limit data to obtain a restriction sub-model. Thus, through the loss sub-model and the restriction sub-model, the target reward function can comprehensively consider the reward value of resource optimization from the dimensions of transaction loss and transaction restriction. Finally, a model is constructed based on the loss sub-model, the restriction sub-model, the preset risk control sub-model and the preset benefit control sub-model to realize the construction of a candidate reward model that can comprehensively evaluate the reward value of resource configuration from multiple aspects.

[0091] In step S201 of some embodiments, the resource transaction configuration information is the rules for trading the target resource, such as shipping loss rate, holding cost, etc. The resource transaction configuration information can be recorded in text or XML format, and this application does not impose any restrictions. In one embodiment, the resource transaction configuration information is code information, and the preset semantic extraction model is a large-scale natural language model. The code recording the resource transaction configuration information is input into the large-scale natural language model for semantic extraction, thereby obtaining transaction loss data and transaction limit data for the target resource during the transaction.

[0092] In step S202 of some embodiments, a loss model is constructed based on transaction loss data. The transaction loss data is the loss of target resources during transactions. For example, when trading medical device inventory, logistics transportation is required, and thus there is transaction loss data for logistics transportation. The loss model is constructed as shown in formula (1):

[0093] (1),

[0094] in, is the weight coefficient of the loss sub-model, which is the preset value when constructing the loss sub-model. Transaction loss data.

[0095] In step S203 of some embodiments, the transaction number limit is a daily transaction number limit. For example, in medical inventory, the number of outbound configurations can be limited, the number of inbound configurations can be limited, or the number of configurations can be limited. This application does not impose specific restrictions. The restriction model is constructed as shown in formula (2):

[0096] (2),

[0097] in, is the weight coefficient of the restricted sub-model, which is the preset value when constructing the restricted sub-model. Limit data for the number of transactions.

[0098] In step S204 of some embodiments, the preset risk control sub-model is a sub-model for controlling the target resource to adopt an extreme resource allocation strategy. For example, when the value of medical devices fluctuates greatly, in order to avoid risks, risk aversion will be considered and a conservative inventory allocation strategy will be adopted. The preset benefit control sub-model is for controlling the value benefits generated by the target resource in resource allocation. For example, when the value of medical devices is trending upward, timely external allocation of medical devices will generate value benefits. The candidate reward model obtained by model construction is shown in formula (3):

[0099] (3),

[0100] in, is the weight coefficient of the preset benefit control sub-model, The preset benefit control sub-model includes the maximum drawdown sub-model and the risk exposure sub-model. is the weight coefficient of the maximum drawdown sub-model, is the maximum drawdown sub-model, is the weight coefficient of the risk exposure sub-model, is the candidate reward model.

[0101] See also Figure 3 In some embodiments, step S103 may include but is not limited to steps S301 to S307:

[0102] Step S301, chromosome encoding is performed on the candidate reward model to obtain an initial function population;

[0103] Step S302: Screen the initial function population for fitness based on historical transaction data to obtain a first function population;

[0104] Step S303, performing crossover and mutation processing on the initial function population to obtain a second function population;

[0105] Step S304: performing a local search based on the second function population and historical transaction data to obtain a third function population;

[0106] Step S305, performing population fitness replacement on the second function population according to the first function population to obtain a fourth function population;

[0107] Step S306, merging the third function population and the fourth function population to obtain a merged function population;

[0108] Step S307: Perform fitness screening on the merged function population based on historical transaction data to obtain a target reward function.

[0109] In steps S301 to S307 shown in the embodiment of the present application, the candidate reward model is chromosomally encoded to obtain an initial function population, and then the initial function population is fitness screened according to historical transaction data to obtain a first function population, and then cross-mutation processing is performed on the initial function population to obtain a second function population, and then local search is performed based on the second function population and historical transaction data to obtain a third function population, and then the second function population is replaced with the first function population based on the population fitness to obtain a fourth function population, and then the third function population and the fourth function population are merged to obtain a merged function population, and finally the merged function population is fitness screened according to the historical transaction data, so as to obtain a target reward function that can comprehensively evaluate the loss, transaction number limit, resource allocation risk and resource allocation benefit in resource allocation, thereby improving the accuracy of resource allocation.

[0110] In step S301 of some embodiments, chromosome encoding refers to the process of chromosome encoding the weight coefficients of each model of the candidate reward model. The candidate reward model includes Five weight coefficients, then convert the five weight coefficients into arrays, and then assign random values ​​to the five weight coefficients to obtain multiple groups of arrays, and the multiple groups of arrays constitute the initial function population.

[0111] In step S302 of some embodiments, the initial function population is screened for fitness based on the historical transaction data. First, a fitness calculation is performed on the initial function population within the historical transaction data, and then populations with high fitness are selected to obtain a first function population. In one embodiment, the fitness calculation is performed on the initial function population based on the historical transaction data to obtain the top 10 percent of populations with the highest fitness, thereby obtaining the first function population. It should be noted that the threshold for fitness screening can be set as desired, and this application does not impose any specific numerical restrictions.

[0112] In step S303 of some embodiments, crossover and mutation processing is performed on the initial function population to obtain a second function population. The crossover processing is to perform chromosome crossover on the individuals in the initial function population, and the mutation processing is to randomly mutate the values ​​of the individuals in the initial function population. The population after the crossover processing and the population after the mutation processing are then used as new populations to obtain the second function population.

[0113] In step S304 of some embodiments, a local search is performed based on the second function population and the historical transaction data. The local search is to perform fitness screening on the second function population based on the historical transaction data, screen out individuals whose fitness meets the threshold based on a pre-set local search threshold, and then search based on a pre-set search value to obtain new individuals, i.e., obtain a third function population. In one embodiment, the individuals whose fitness meets the threshold based on the pre-set local search threshold are (3, 2.1, 2, 1, 0.5), the search value set by the cosine is positive 0.1 or negative 0.1, and then, based on each value in (3, 2.1, 2, 1, 0.5), 0.1 is added or subtracted to obtain 32 new individuals, and these 32 new individuals are formed into a new population, i.e., the third function population.

[0114] In step S305 of some embodiments, population fitness replacement involves first calculating the fitness of the second function population based on historical transaction data, then obtaining the number of individuals in the first function population to obtain the number of replacements, and then replacing the individuals in the second function population corresponding to the reciprocal number of replacements in fitness with individuals in the first function population. For example, if there are five individuals in the first function population, then the individuals with the reciprocal fitness values ​​in the second function population are replaced with individuals in the first function population to obtain a fourth function population.

[0115] In step S306 of some embodiments, the third function population and the fourth function population are merged. When merging the populations, the third function population and the fourth function population are first merged, and then the merged population is removed for deduplication to obtain a merged function population.

[0116] See also Figure 4 In some embodiments, step S307 may include but is not limited to steps S401 to S404:

[0117] Step S401, calculating the reward value of the merged function population based on historical transaction data to obtain target reward value data;

[0118] Step S402: Allocate resources to the historical transaction data based on the reward value data to obtain resource income data and resource withdrawal value data;

[0119] Step S403, performing fitness calculation based on the resource income data and the resource withdrawal value data to obtain population fitness data;

[0120] Step S404: Screen the merged function population according to the population fitness data to obtain the target reward function.

[0121] In steps S401 to S404 shown in the embodiment of the present application, reward values ​​are calculated for the merged function population based on historical transaction data to obtain target reward value data, and then resources are allocated to the historical transaction data based on the reward value data to obtain resource income data and resource withdrawal value data, and then fitness is calculated based on the resource income data and resource withdrawal value data to obtain population fitness data, and finally, the merged function population is screened based on the population fitness data to obtain the target reward function, thereby obtaining a reward function that can perform reasonable resource allocation evaluation, thereby improving the accuracy of resource allocation.

[0122] See also Figure 5 In some embodiments, step S401 includes but is not limited to steps S501 to S504:

[0123] Step S501, calculating the resource retention reward value for the merged function population based on historical transaction data to obtain resource retention reward value data;

[0124] Step S502: Calculate the resource output reward value of the merged function population based on the historical transaction data to obtain resource output reward value data;

[0125] Step S503, calculating the resource input reward value for the merged function population based on the historical transaction data to obtain resource input reward value data;

[0126] Step S504: taking the resource retention reward value data, the resource output reward value data, and the resource input reward value data as target reward value data.

[0127] In steps S501 to S504 shown in the embodiment of the present application, the resource retention reward value is calculated for the merged function population based on the historical transaction data to obtain resource retention reward value data, and then the resource output reward value is calculated for the merged function population based on the historical transaction data to obtain resource output reward value data, and the resource input reward value is calculated for the merged function population based on the historical transaction data to obtain resource input reward value data. Finally, the resource retention reward value data, resource output reward value data and resource input reward value data are used as target reward value data.

[0128] In step S501 of some embodiments, the historical transaction data includes the value data of the target resource in multiple time periods. For example, on the first day, the resource value is the first value, and on the second day, the resource value is the second value. On the first day, a decision is made based on the allocation decision of resource reservation, that is, the medical device inventory is reserved. Then, the reward value in the merged function population is calculated based on the first value on the first day and the second value on the second day. The resource reservation reward value is calculated as shown in formula (4):

[0129] (5),

[0130] in, Keep reward value data for resources, Allocation decisions for resource reservations, is the resource value on a specific day. For example, on the first day, the resource value is the first value, i.e. is the first value, and on the second day, the resource value is the second value, that is, As the second value. Then based on resource reservation allocation decision, calculate the benefit data after resource reservation , maximum drawdown , risk exposure , transaction loss data , transaction limit data When making allocation decisions for resource conservation, benefit data 0, transaction loss data 0, transaction limit data is 0, the maximum drawdown The maximum drawdown value calculated based on the allocation strategy of historical value data and resource reservation, risk exposure The risk exposure value is calculated based on historical value data and resource output allocation strategy.

[0131] In step S502 of some embodiments, the resource reservation reward value is calculated as shown in formula (5):

[0132] (5),

[0133] in, For resource output allocation decisions, Outputs reward value data for a resource.

[0134] In step S503 of some embodiments, the resource input reward value is calculated as shown in formula (6):

[0135] (6),

[0136] in, Allocation decisions for resource inputs, Enter reward value data for the resource.

[0137] In step S504 of some embodiments, the resource retention reward value data, the resource output reward value data and the resource input reward value data are used as target reward value data. The target reward value data is in the form of , this application does not impose any specific restrictions on the form of the target reward value data.

[0138] In step S402 of some embodiments, the reward value data includes resource retention reward value data, resource output reward value data, and resource input reward value data. The largest data is selected from the reward value data as the resource allocation decision. For example, if the largest data in the reward value data is the resource output reward value data, the resource output is used as the resource allocation decision. The historical transaction data includes the value data of the target resource in multiple time periods. Based on the merge function population, a resource allocation decision is made in each time period, and the current resource income data and the current resource drawdown value data are calculated based on the next time period. The cycle is repeated until a resource allocation decision is made in each time period until the historical transaction data for each time period is calculated, and finally the overall resource income data and resource drawdown value data are calculated.

[0139] In step S403 of some embodiments, fitness calculation is performed based on the resource benefit data and the resource withdrawal value data. The fitness calculation is shown in formula (7):

[0140] (7),

[0141] in, is the fitness data, is the resource income data, is the preset resource income data weight, It is the resource withdrawal value data. The data weight of the resource withdrawal value.

[0142] In step S404 of some embodiments, the function population is merged based on the population fitness data, and the individual with the highest fitness is selected as the target reward function.

[0143] See also Figure 6 In some embodiments, the preset decision generation model includes experience samples, an initial decision layer, and an initial prediction layer; the experience samples include an execution action, a current state, and a next state. Step S104 includes but is not limited to steps S601 to S605:

[0144] Step S601: Perform decision evaluation based on the execution action, current state, and initial decision layer to obtain initial decision evaluation data; wherein the initial decision evaluation data is used to indicate the rationality of the decision made by the initial decision layer;

[0145] Step S602: performing a prediction evaluation based on the execution action, the next state, and the initial prediction layer to obtain initial prediction evaluation data; wherein the initial prediction evaluation data is used to indicate the rationality of the decision of the initial prediction layer;

[0146] Step S603: Calculate the reward value based on the historical transaction data, current state, executed action, next state, and target reward function to obtain a training reward value;

[0147] Step S604, optimizing the parameters of the initial decision layer according to the training reward value, the initial prediction evaluation data, and the initial decision evaluation data to obtain a target decision layer;

[0148] Step S605: Generate a target decision model based on the experience sample, the target decision layer and the initial prediction layer.

[0149] In steps S601 to S605 shown in the embodiment of the present application, initial decision evaluation data is obtained by performing decision evaluation based on the execution action, current state and initial decision layer, and then prediction evaluation is performed based on the execution action, next state and initial prediction layer to obtain initial prediction evaluation data, reward value calculation is performed based on historical transaction data, current state, execution action, next state and target reward function to obtain training reward value, and then parameters of the initial decision layer are optimized based on the training reward value, initial prediction evaluation data and initial decision evaluation data, so that the initial decision layer can perceive the next state under the current state and predict further states based on the next state to obtain the target decision layer, and finally, the experience sample, target decision layer and initial prediction layer are used as the target decision generation model to obtain a target decision generation model that can accurately make resource allocation decisions.

[0150] In step S601 of some embodiments, the preset decision generation model is a deep Q network (DQN). DQN includes experience samples, an initial decision layer and an initial prediction layer. The initial decision layer and the experience sample are a buffer pool that stores execution actions, current states and next states. The execution actions include three actions: resource output, resource input and resource retention. The current state is the total value of the current medical device inventory, and the next state is the total value of the medical device inventory after the next state is executed based on the current state. The initial decision layer and the initial prediction layer are both composed of fully connected layers. First, an experience sample is extracted from the prediction decision generation model, and the current state, that is, the total value of the current medical device inventory, is obtained from the experience sample. The execution action is obtained from the experience sample. The execution action may be resource input, resource retention or resource output. The next state is the state after the execution action is performed based on the current state. For example: the experience sample is (current state: total value of medical device inventory: 100, execution action: retain, next state is medical device inventory value 105), and then a decision evaluation is performed based on the current state and the initial decision layer to obtain the evaluation data for executing the retain action in the current state.

[0151] In step S602 of some embodiments, in one embodiment, the experience sample is (current state: total medical device inventory value: 100, execution action: reserve, next state is medical device inventory value 105), and according to the initial prediction layer, in the next state, that is, when the total medical device inventory value is 105, the decision evaluation values ​​of the three execution actions of resource reservation, resource input and resource output are predicted, and the largest decision evaluation value is selected as the initial prediction evaluation data.

[0152] In step S603 of some embodiments, a reward value is calculated based on historical transaction data, current state, execution action, next state and target reward function to obtain a training reward value, that is, the benefit is calculated based on the current state and the next state, the loss is calculated based on the execution action, the number of transaction limits is calculated based on the execution action, and the maximum drawdown value and risk exposure are calculated based on historical transaction data, the current state and the next state. Finally, the training reward value is calculated by combining the benefit, loss, number of transaction limits, maximum drawdown value, risk exposure and target reward function.

[0153] See also Figure 7 In some embodiments, step S604 may include but is not limited to steps S701 to S703:

[0154] Step S701, performing aggregation calculation based on the training reward value and the initial prediction evaluation data to obtain the training label value;

[0155] Step S702: Calculate the mean square error based on the training label value and the initial decision evaluation data to obtain a training loss value;

[0156] Step S703: Optimize the parameters of the initial decision layer according to the training loss value to obtain the target decision layer.

[0157] In steps S701 to S703 shown in the embodiment of the present application, aggregation calculations are performed based on the training reward value and the initial prediction evaluation data to obtain a training label value, and then a mean square error calculation is performed based on the training label value and the initial decision evaluation data to obtain a training loss value. Finally, the parameters of the initial decision layer are optimized based on the training loss value to obtain a target decision layer, so that the target decision layer can have a good perception of resource allocation decisions, thereby making accurate resource allocation decisions.

[0158] In step S701 of some embodiments, an aggregate calculation is performed based on the training reward value and the initial prediction evaluation data as shown in formula (8):

[0159] (8),

[0160] in, is the training reward value, Evaluate data weights for a preset initial forecast, is the initial prediction data, and the training label value is .

[0161] In step S702 of some embodiments, the mean square error is calculated based on the training label value and the initial decision evaluation data as shown in formula (9):

[0162] (9),

[0163] in, Evaluate data for initial decision making, is the training loss value.

[0164] In step S703 of some embodiments, the parameters of the initial decision layer are optimized based on the training loss value, that is, the initial decision layer is back-propagated through the training loss value to obtain the target decision layer.

[0165] In step S605 of some embodiments, the experience samples, the target decision layer and the initial prediction layer are model merged to obtain a target decision generation model. It should be noted that in one embodiment, the target decision layer and the initial decision layer are network structures of the same dimension and the same structure. When training the network, the decision layer is first trained. After iterating to the preset conditions, the network value of the decision layer is assigned to the prediction layer, and then the decision layer is trained again to finally obtain the target decision layer, the target prediction layer and the experience samples. In the embodiments of the present application, only the training of the decision layer to obtain the target decision layer is described. In other embodiments, the network structure can be trained as needed, and this application does not impose specific restrictions.

[0166] In step S105 of some embodiments, after obtaining the current transaction data of the target resource, the current transaction data is input into the target decision generation model. The target decision generation model outputs the evaluation values ​​of the three execution actions of resource input, resource output, and resource retention. The execution action corresponding to the largest evaluation value is selected as the resource allocation decision, and the target resource allocation data is obtained.

[0167] See also Figure 8 The embodiment of the present application further provides a resource allocation device that can implement the above resource allocation method, and the device includes:

[0168] The data acquisition module 801 is used to acquire resource transaction configuration information of the target resource and acquire historical transaction data of the target resource;

[0169] A model acquisition module 802 is used to acquire a candidate reward model based on resource transaction configuration information;

[0170] Genetic search module 803, for performing genetic search based on candidate reward models and historical transaction data to obtain a target reward function;

[0171] A model training module 804 is configured to train a preset decision generation model based on a target reward function and historical transaction data to obtain a target decision generation model;

[0172] The allocation decision module 805 is used to obtain the current resource transaction data of the target resource, and make an allocation decision on the current resource transaction data based on the target decision generation model to obtain the target resource allocation data.

[0173] The specific implementation of the resource allocation device is substantially the same as the specific embodiment of the resource allocation method described above, and will not be described in detail here.

[0174] The present application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the resource allocation method described above when executing the computer program. The electronic device can be any smart terminal, such as a tablet computer or an in-vehicle computer.

[0175] See also Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0176] The processor 901 can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0177] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called by the processor 901 to execute the resource allocation method of the embodiments of this application.

[0178] Input / output interface 903, used to implement information input and output;

[0179] Communication interface 904, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0180] Bus 905 , which transmits information between various components of the device (e.g., processor 901 , memory 902 , input / output interface 903 , and communication interface 904 );

[0181] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0182] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned resource allocation method is implemented.

[0183] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0184] The resource allocation method, resource allocation device, electronic device and storage medium provided in the embodiments of the present application obtain resource transaction configuration information and historical transaction data of the target resource, obtain a candidate reward model based on the resource transaction configuration information, and perform genetic search based on the reward function model and historical transaction data to obtain a target reward function. Then, a preset decision generation model is trained based on the target reward function and historical transaction data, so that the preset decision generation model can be trained into a target decision generation model that can make resource allocation decisions with high allocation efficiency under the reward incentive of the target reward function, thereby improving the efficiency and accuracy of resource allocation; further, the current resource transaction data of the target resource is obtained, and an allocation decision is made on the current resource transaction data based on the target decision generation model to obtain target resource allocation data, thereby achieving efficient and accurate resource allocation.

[0185] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0186] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0187] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0188] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0189] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0190] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0191] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0192] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0193] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0194] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0195] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A resource allocation method, characterized in that: The method comprises: Obtain resource transaction configuration information of a target resource, and obtain historical transaction data of the target resource; wherein the target resource is a medical device resource; Acquire a candidate reward model based on the resource transaction configuration information; wherein the resource transaction configuration information includes transaction loss data and transaction number limit data, and the candidate reward model includes a loss sub-model, a limit sub-model, a preset risk control sub-model, and a preset benefit control sub-model; Perform chromosome encoding on the candidate reward model to obtain an initial function population; Performing fitness screening on the initial function population according to the historical transaction data to obtain a first function population; Performing crossover and mutation processing on the initial function population to obtain a second function population; Performing a local search based on the second function population and the historical transaction data to obtain a third function population; Performing population fitness replacement on the second function population according to the first function population to obtain a fourth function population; Merging the third function population and the fourth function population to obtain a merged function population; Calculating the reward value of the merged function population according to the historical transaction data to obtain target reward value data; Allocate resources to the historical transaction data according to the reward value data to obtain resource income data and resource withdrawal value data; Performing fitness calculation based on the resource income data and the resource withdrawal value data to obtain population fitness data; Screening the merged function population according to the population fitness data to obtain a target reward function; Training a preset decision generation model based on the target reward function and the historical transaction data to obtain a target decision generation model; Current resource transaction data of the target resource is acquired, and allocation decision is made on the current resource transaction data based on the target decision generation model to obtain target resource allocation data.

2. The method according to claim 1, characterized in that The step of calculating the reward value of the merged function population according to the historical transaction data to obtain target reward value data includes: Calculating a resource retention reward value for the merged function population according to the historical transaction data to obtain resource retention reward value data; Calculating the resource output reward value of the merged function population according to the historical transaction data to obtain resource output reward value data; Calculating the resource input reward value for the merged function population according to the historical transaction data to obtain resource input reward value data; The resource retention bonus value data, the resource output bonus value data, and the resource input bonus value data are used as the target bonus value data.

3. The method according to any one of claims 1 to 2, characterized in that The preset decision generation model includes experience samples, an initial decision layer, and an initial prediction layer; the experience samples include an execution action, a current state, and a next state; the preset decision generation model is trained based on the target reward function and the historical transaction data to obtain a target decision generation model, including: Performing a decision evaluation based on the execution action, the current state, and the initial decision layer to obtain initial decision evaluation data; wherein the initial decision evaluation data is used to indicate the rationality of the decision of the initial decision layer; Performing a prediction evaluation based on the execution action, the next state, and the initial prediction layer to obtain initial prediction evaluation data; wherein the initial prediction evaluation data is used to indicate the rationality of the decision of the initial prediction layer; Calculating a reward value based on the historical transaction data, the current state, the executed action, the next state, and the target reward function to obtain a training reward value; Optimizing parameters of the initial decision layer according to the training reward value, the initial prediction evaluation data, and the initial decision evaluation data to obtain a target decision layer; The target decision generation model is formed according to the experience samples, the target decision layer and the initial prediction layer.

4. The method according to claim 3, characterized in that The step of optimizing parameters of the initial decision layer according to the training reward value, the initial prediction evaluation data, and the initial decision evaluation data to obtain a target decision layer includes: Performing aggregate calculation based on the training reward value and the initial prediction evaluation data to obtain a training label value; Calculate the mean square error based on the training label value and the initial decision evaluation data to obtain a training loss value; Parameters of the initial decision layer are optimized according to the training loss value to obtain the target decision layer.

5. The method according to any one of claims 1 to 2, characterized in that The acquiring of a candidate reward model based on the resource transaction configuration information includes: Perform semantic extraction on the resource transaction configuration information according to a preset semantic extraction model to obtain transaction loss data and transaction number limit data; Constructing a loss model based on the transaction loss data to obtain a loss sub-model; Constructing a restriction model based on the transaction number restriction data to obtain a restriction sub-model; A model is constructed based on the loss sub-model, the restriction sub-model, the preset risk control sub-model and the preset benefit control sub-model to obtain the candidate reward model.

6. A resource allocation device, characterized in that: The device comprises: A data acquisition module is used to acquire resource transaction configuration information of a target resource and acquire historical transaction data of the target resource; wherein the target resource is a medical device resource; A model acquisition module is configured to acquire a candidate reward model based on the resource transaction configuration information; wherein the resource transaction configuration information includes transaction loss data and transaction number limit data, and the candidate reward model includes a loss sub-model, the limit sub-model, a preset risk control sub-model, and a preset benefit control sub-model; A genetic search module, configured to perform chromosome encoding on the candidate reward model to obtain an initial function population; Performing fitness screening on the initial function population according to the historical transaction data to obtain a first function population; Performing crossover and mutation processing on the initial function population to obtain a second function population; Performing a local search based on the second function population and the historical transaction data to obtain a third function population; Performing population fitness replacement on the second function population according to the first function population to obtain a fourth function population; Merging the third function population and the fourth function population to obtain a merged function population; Calculating the reward value of the merged function population according to the historical transaction data to obtain target reward value data; Allocate resources to the historical transaction data according to the reward value data to obtain resource income data and resource withdrawal value data; Performing fitness calculation based on the resource income data and the resource withdrawal value data to obtain population fitness data; Screening the merged function population according to the population fitness data to obtain a target reward function; A model training module, configured to train a preset decision generation model based on the target reward function and the historical transaction data to obtain a target decision generation model; The allocation decision module is used to obtain the current resource transaction data of the target resource and make an allocation decision on the current resource transaction data based on the target decision generation model to obtain the target resource allocation data.

7. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the resource allocation method according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the resource allocation method according to any one of claims 1 to 5 is implemented.

Images